As the self scheduling node described in protocol-spool and protocol-multigraph will make it much more practical to have munin graph network traffic on all ports of a switch - indeed on all switches in a organization. That creates too many graphs at the top level of each monitored node.

A plugin should be able to name the datapoints in a hierarchic way and have the graphs be organized in the same hierarchy. The plugin should also describe the top-level graph in some way - one example for switch port load would be that the toplevel graph shows total output of all the ports on the switch.